Blind separation of audio signal mixtures

• Main Author: Ng, Hock Seng.
• Other Authors: Wan, Chunru
• Format: Theses and Dissertations
• Published: 2008
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
• Online Access: http://hdl.handle.net/10356/4948
• Institution: Nanyang Technological University

The separation of two unknown signals that mixed together in an unknown manner was shown to be possible. This process is called blind source separation. Two types of mixing process, which are instantaneous and convolutive mixing were simulated. It was found that by assuming that the two known signals and their signal mixtures were uncorrelated, the two original signals could be recovered. The signals undergoing instantaneous mixing can be easily separated by eigen-decomposition method processing. The signals undergoing convolutive mixing, however need a more robust separation algorithm. They still can be separated using gradient-based optimisation algorithm and simulated annealing algorithm, which generates the solution signals set that correspond to the least cross-correlation value.